使用zookeeper管理多个hbase集群

阅读更多

    zookeeper是hbase集群的"协调器"。由于zookeeper的轻量级特性，因此我们可以将多个hbase集群共用一个zookeeper集群，以节约大量的服务器。多个hbase集群共用zookeeper集群的方法是使用同一组ip，修改不同hbase集群的"zookeeper.znode.parent"属性，让它们使用不同的根目录。比如cluster1使用/hbase-c1,cluster2使用/hbase-c2，等等。
 
    使用以上方法有一个现实的问题：如何避免各集群的相互干扰？因为client的配置权是在用户手上，并不能保证用户永远是配置正确的，那么会产生某个用户访问了不该他访问的hbase集群。此时数据安全性成了很大的问题，甚至可能出现误删除数据。我们需要在zookeeper层屏弊掉该问题。
 
    zookeeper3.x版本起自带了简单的ACL功能(注意3.3.x版本起不再支持按hostname来分配权限)。见： http://zookeeper.apache.org/doc/r3.3.2/zookeeperProgrammers.html#sc_ZooKeeperAccessControl。进行权限配置主要使用digest和ip两种方法。其中digest是用户密码方式，对用户来说使用上并不透明。ip配置最简单，对用户也是透明的，用户并不知道的情况下就能限制它的访问权限。
 
    zookeeper将访问权限分为了五类:READ/WRITE/DELETE/CREATE/ADMIN，其中admin为最高权限。zookeeper的权限是到znode级别的，限制了某一个node的权限并不能限制它的子节点权限。

    不过使用IP做权限配置方案有一个缺陷：必须指定具体的ip，而不能使用通配符或者范围一类的。这样对于大规模的权限设置是非常不方便的一件事，因此作者略调整了一下zookeeper的代码：
  IPAuthenticationProvider.java
 

/**
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

package org.apache.zookeeper.server.auth;

import org.apache.zookeeper.data.Id;
import org.apache.zookeeper.server.ServerCnxn;
import org.apache.zookeeper.KeeperException;

public class IPAuthenticationProvider implements AuthenticationProvider {

    public String getScheme() {
        return "ip";
    }

    public KeeperException.Code
        handleAuthentication(ServerCnxn cnxn, byte[] authData)
    {
        String id = cnxn.getRemoteAddress().getAddress().getHostAddress();
        cnxn.getAuthInfo().add(new Id(getScheme(), id));
        return KeeperException.Code.OK;
    }

    // This is a bit weird but we need to return the address and the number of
    // bytes (to distinguish between IPv4 and IPv6
    private byte[] addr2Bytes(String addr) {
        byte b[] = v4addr2Bytes1(addr);
        // TODO Write the v6addr2Bytes
        return b;
    }

    private byte v4addr2Bytes(String part) throws NumberFormatException{
        try {
           int v = Integer.parseInt(part);
           if (v >= 0 && v <= 255) {
                  byte  b = (byte) v;
                  return b;
                } else {
                    throw new NumberFormatException("v < 0 or v > 255!");
                }
            } catch (NumberFormatException e) {
                throw e;
            }
    }
    
    private byte[] v4addr2Bytes1(String addr) {
      String parts[] = addr.split("\\.", -1);
      if (parts.length != 4) {
          return null;
      }
      byte b[] = new byte[4];
      for (int i = 0; i < 4; i++) {
      	try {
      		if(parts[i].split("/").length == 2){
      			v4addr2Bytes(parts[i].split("/")[0]);
      			v4addr2Bytes(parts[i].split("/")[1]);
      			continue;
      		}else{
            b[i] = v4addr2Bytes(parts[i]);
      		}
      	} catch (NumberFormatException e) {
          return null;
  			}
      }
      return b;
  }
    
    public boolean matches(String id, String aclExpr) {
      String parts[] = aclExpr.split("/", 2);
      byte aclAddr[] = addr2Bytes(parts[0]);
      if (aclAddr == null) {
          return false;
      }
      byte endAclAddr[] = new byte[aclAddr.length];
      for(int i = 0; i < aclAddr.length; i ++){
      	endAclAddr[i] = aclAddr[i];
      }
      if (parts.length == 2) {
          try {
          		int end = Integer.parseInt(parts[1]);
          		int e = endAclAddr[endAclAddr.length-1]<=0?endAclAddr[endAclAddr.length-1]+256:endAclAddr[endAclAddr.length-1];
          		if(end <  e|| end < 0 || end > 255)
          			return false;
              endAclAddr[endAclAddr.length-1] = (byte)end;
          } catch (NumberFormatException e) {
              return false;
          }
      }
      byte remoteAddr[] = addr2Bytes(id);
      if (remoteAddr == null) {
          return false;
      }
      for (int i = 0; i < remoteAddr.length; i++) {
      	int r = remoteAddr[i]<=0?(int)remoteAddr[i]+256:remoteAddr[i];
      	int a = aclAddr[i]<=0?(int)aclAddr[i]+256:aclAddr[i];
      	int e = endAclAddr[i]<=0?(int)endAclAddr[i]+256:endAclAddr[i];
        if (r < a || r > e) {
          return false;
        }
      }
      return true;
  }

    public boolean isAuthenticated() {
        return false;
    }

    public boolean isValid(String id) {
        return addr2Bytes(id) != null;
    }
}

    支持了使用/做为范围标识，比如进入hbase zkcli，执行：setAcl /test ip:192.168.0.3/10:cd，则将读写权限赋给了192.168.0.3-192.168.0.10这8台机器，其它机器将没有任何权限。
    这样用同一个zookeeper管理多个集群、海量机器将不再有困扰。
    最后写了一个帮助运维同学自动化管理zookeeper集群下多个hbase集群的ACL权限的工具，像以下这样：

java -Djava.ext.dirs=libs/ -cp hbase-tools.jar dwbasis.hbase.tools.client.ZookeeperAcl aclFile.json
Usage: ZookeeperAcl acljsonfile [-plan]
/test/t ==> 'ip,'192.168.0.1
:cdrwa
/test ==> 'ip,'192.168.0.1/3
:cdrwa
/test ==> 'ip,'192.168.0.5
:cdrwa
do you really setAcl as above?(y/n)

    补充：多集群共用zk后，每个集群的启动和停止不应该影响zk的稳定。因此请配置hbase-env.sh中export HBASE_MANAGES_ZK=false